Related Experiment Video
Updated: Aug 8, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Exploring the Connection between Uncertainty and Tissue Boundaries in Medical Image Segmentation
IEEE Transactions on Medical Imaging
|August 6, 2026
Summary
This study introduces the Evidential Uncertainty-Guided Boundary (EUGB) loss to improve medical image segmentation accuracy, particularly at blurry boundaries. The novel loss function enhances boundary delineation by leveraging uncertainty information for better computer-aided diagnosis.
Area of Science:
- Medical image analysis
- Computer-aided diagnosis
- Machine learning for healthcare
Background:
- Automatic medical image segmentation is crucial for quantitative analysis and computer-aided diagnosis.
- Blurry boundaries in medical images, due to imaging quality or tissue properties, lead to imprecise segmentation.
- High uncertainty and misclassification are common at object boundaries during segmentation.
Purpose of the Study:
- To investigate the relationship between uncertainty and tissue boundaries in medical image segmentation.
- To propose a novel loss function, Evidential Uncertainty-Guided Boundary (EUGB) loss, to address boundary segmentation errors.
- To demonstrate that incorporating uncertainty information can improve segmentation performance.
Main Methods:
- Exploration of uncertainty and tissue boundary connections across various tissues.
- Development and implementation of the Evidential Uncertainty-Guided Boundary (EUGB) loss function.
- Validation of EUGB loss on LIDC-IDRI, ISIC 2018, and OCTA-500 datasets using U-Net and TransU-Net models.
Main Results:
- The proposed EUGB loss effectively emphasizes challenging pixels along blurry boundaries using evidential uncertainty.
- EUGB loss incorporates a regularization term to constrain uncertainty learning, penalizing incorrect predictions and reinforcing correct ones.
- Experimental results show EUGB loss outperforms seven other segmentation loss functions in boundary segmentation accuracy while maintaining competitive region-level accuracy.
Conclusions:
- The EUGB loss function significantly improves boundary segmentation in medical images by effectively utilizing uncertainty information.
- The study provides empirical insights into selecting appropriate loss functions based on dataset characteristics and application scenarios.
- This work offers practical guidance for researchers and practitioners to optimize medical image segmentation performance.
